ResearchBenchmarksNon-Linear Reward Scalarization on Cancer_nl (held-out settings)Follow91.94RewardKθ − Û40.87654.13367.3980.647Mar 2, 2026Evaluation ResultsMethodMethodLinksRewardKθ − Ûkernel=Learned (Kθ), H...kernel=Learned (Kθ), HTI_component=Included2026.0391.94KI − Ûkernel=Identity (KI),...kernel=Identity (KI), HTI_component=Included2026.0386.21MFM2026.0378.44CFM2026.0369.7Kθkernel=Learned (Kθ)kernel=Learned (Kθ)2026.0354.44NLOT2026.0352.63Direct2026.0349.5KIkernel=Identity (KI)kernel=Identity (KI)2026.0342.84